Mission control for AI agents.

Grain is where your team orchestrates, reviews, and ships work from autonomous agents. One surface for tasks, diffs, approvals, and handoffs — web, desktop, Telegram, or Slack.

Download for Mac Open in browser

grain fleet

# Spin up a fleet from a single prompt
$ grain fleet "build checkout v2: Stripe + Apple Pay, a11y AA, Playwright tests"

✔ planning… 4 linked tasks created
✔ agent-1 → api/checkout   (claude-sonnet · task-a3f9)
✔ agent-2 → ui/cart         (codex · task-b12e)
✔ agent-3 → tests/checkout  (claude-haiku · task-c77d)
✔ agent-4 → docs            (gemini-cli · task-d04b)

Live at https://app.grain.sh/fleet/0xgrn…

20+

connected models & CLIs (Claude Code, Codex, OpenCode, Gemini CLI)

10s

live diff stream — see changes as the agent types

channels — Web, Desktop, Telegram, Slack, Discord, Teams

Download

native app for Mac, Linux & Windows

How it works

Three steps from first prompt to production.

Connect your fleet

Bring your own models: Claude Code, Codex, OpenCode, Gemini CLI, or API keys. Grain runs a local executor bundled in a Tauri desktop app — 74MB, no Node.js required.

Prompt the fleet

Ship one prompt and Agentic Mode plans the work across parallel linked tasks. Or assign a single task to a single agent. Task history, costs, and iterations live in one place.

Review and ship

Live diffs stream every 10 seconds. Leave inline comments, request changes, approve, or hand off to a teammate. Grain pushes the final commit when you're done.

Built for every seat

However you go to market, the agent adapts.

Engineering leaders

Give your team one place to oversee every coding agent — who's running what, where diffs live, what needs review. Replace a tab soup of CLIs with a command center.

Platform teams

Self-host Grain on your Convex instance. Bring your own model providers. Define org-wide Skills for coding conventions, security review, and release checklists.

Solo founders

Run a fleet from your laptop. Open a prompt on the desktop app in the morning, approve pull requests from Telegram at lunch, ship by end of day.

Agencies & consultancies

Keep every client in its own workspace. Share task links with non-technical stakeholders. Ship 3x the work with the same headcount.

Plugs into your stack

Native integrations across the tools your team already uses.

Claude CodeCodexOpenCodeGemini CLIGitHubGitLabLinearJiraNotionTelegramSlackDiscordTeamsConvexVercel

Frequently asked

How is Grain different from running Claude Code in a terminal?

A terminal gives you one agent, one shell, one machine. Grain gives you a fleet — many agents, many tasks in flight, a review queue, live diffs, inline comments, multi-channel notifications, and a shared Skills library your whole team inherits. Same raw models underneath; the difference is the team surface.

Which models can I run?

Anything with a CLI or API: Claude Code, Codex, OpenCode, Gemini CLI, Ollama, vLLM, or direct provider APIs (Anthropic, OpenAI, Google, DeepSeek, Qwen). You bring the keys, Grain routes the work.

Is Grain self-hostable?

Yes. Grain runs on self-hosted Convex (ports 3210/3211), the Tauri desktop app ships with a bundled local executor, and the web app can be hosted on your own infrastructure. Zero SaaS lock-in.

What about security?

Per-task permission modes, signed commits, full audit log of every agent action, and approval gates on anything that touches production. The executor runs locally by default — code never leaves your machine unless you approve it.

How does pricing work?

Grain itself is free while in preview. You pay your model providers directly — we don't mark up inference. Enterprise pricing covers SSO, priority support, and managed hosting when you're ready.

Download Grain.

Free native app for Mac, Linux, and Windows. Self-host when you're ready.

Download for Mac Linux & Windows